Excellent work and thanks for sharing code. I am wondering does this method work well when applying to in-domain few shot tasks? It means, if I have only one dataset and I just label 10 instances, then train the model and predict the rest of the dataset.
Excellent work and thanks for sharing code. I am wondering does this method work well when applying to in-domain few shot tasks? It means, if I have only one dataset and I just label 10 instances, then train the model and predict the rest of the dataset.